The Perceptual Gap: Why We Need Accessible XAI for Assistive Technologies

AI systems serving people with disabilities lack transparency: The paper identifies a critical gap where assistive AI technologies—such as image description tools for blind users or speech captioning for deaf users—operate as "black boxes" without explanations, meaning disabled users cannot understand why the system made particular decisions about their information.

Current explainability methods are not designed for accessibility: Existing XAI approaches rarely account for sensory disabilities, and many standard explanation formats are difficult or impossible for people with vision or hearing loss to comprehend, creating a secondary barrier to understanding AI decisions that affect them.

Disabled users are underrepresented in XAI research: The survey reveals there is "next-to-no XAI work that accounts for people with sensory disabilities," indicating that accessibility considerations have been largely absent from the explainable AI research agenda despite these populations being significant users of AI-powered assistive technologies.

Accessible explanations require intentional design: The paper suggests that making XAI work for disabled users is not simply a matter of applying existing methods, but requires human-centered and accessibility-centered approaches built from the ground up—implying that responsible AI development must actively include disabled communities in design processes.

This gap has real consequences for trust and autonomy: When assistive AI systems cannot be understood by their users, it undermines disabled individuals' ability to verify accuracy, catch errors, and maintain agency over decisions that directly affect their access to information and independence.